Temporal anaphora in a tenseless language*
Jürgen Bohnemeyer
1. Introduction
This chapter presents a portrait of a language that arguably lacks absolute
(i.e., deictic) and relative (i.e., anaphoric) tenses and temporal connectives
with meanings comparable to those of English after, before, until, and while.
The language is Yucatec Maya. “Tenselessness”, the absence of tenses
from the grammar of a language, has been documented for a number of
languages.1 Yucatec goes beyond tenselessness in its simultaneous lack of
temporal connectives of the indicated kind. However, the emphasis in this
chapter is on tenselessness and on the question how Yucatec speakers manage to communicate about time in the face of it.
It is assumed in this chapter that tenses express binary ordering relations
between the time about which an utterance makes a statement, asks a ques*
1
Sections 2 and 3 of this chapter present a summary of chapters 4 –7 of Bohnemeyer (1998b) and (2002). The research presented in these works was fully
funded by the Max Planck Society. The examples used in section 3 are mostly
new, though, and the analysis of one of the aspect-mood markers of Yucatec –
the remote future (“predictive” in Bohnemeyer 1998b, 2002) – has been revised
in light of new evidence. The account of temporal anaphora in section 4 is an informal version of the analysis I presented at the SULA 5 conference in São Paulo
in May 2007 (though using different material for illustration). This analysis is an
update of the one in Bohnemeyer (1998b, 2000a/b, 2002), preserving its Gricean
core, but attempting a simpler, more concise, and more rigorous formulation (and
integration into the DRT framework, which is not discussed in the present paper).
Cf., e.g., Bittner (2005, 2007) and Shaer (2003) on Kalaallisut (or West Greenlandic); Comrie (1976: 82–84) on Igbo and Yoruba; Comrie (1985: 50–53) on
Burmese and Dyirbal; and Li & Thompson (1981: 184, 213–215) on Mandarin.
Tenselessness may also be discussed in terms of the absence of tense marking in
particular utterances, rather than in the entire grammar. For instance, Smith,
Perkins & Fernald (2007) argue that Navajo has a future-non-future tense system,
but that tense marking is optional in this language. The interpretation of tenseless
utterances in Navajo, based on these authors’ observations, appears to rely on
principles similar to those proposed for Yucatec in section 4.
2 Jürgen Bohnemeyer
tion, or issues a command, etc. – the topic time of the utterance, following
Klein 1994 (see also chapter 2 of this book) – and, in the case of deictic
tense, the time at which the utterance is made or interpreted (the coding
time), or, in the case of anaphoric tenses and temporal connectives, some
time mentioned in discourse – the reference point.2 In this framework, Yucatec can be characterized as a language in which the topic times of utterances are not constrained vis-à-vis utterance times or reference points by
the morphosyntactic form of the clause (but adverbials may of course be
used to determine them). This claim is defended in section 3 and possible
exceptions are discussed there as well. Beyond making the case for the existence of languages such as Yucatec, the main concern of this chapter is the
question of how Yucatec speakers determine the topic times of utterances
in the absence of explicit coding. The proposal developed in section 4 is that
they rely on inference mechanisms of temporal anaphora which are shared
with Indo-European languages and, presumably, universal.
Temporal anaphora is the contextual determination of topic times. Temporal anaphora resolution depends on many factors, including the semantics
of aspectual and modal operators (aside from lexical semantic properties,
rhetorical structure, and world knowledge). The main emphasis in the present chapter is on the impact of aspectual operators. Consider (1):
(1)
a. Floyd entered. Sally made a phone call.
b. Floyd entered. Sally was making a phone call.
According to the standard analysis of temporal anaphora in Discourse Representation Theory (DRT; cf. Kamp 1979; Kamp & Rohrer 1983; Kamp &
Reyle 1993; Hinrichs 1981, 1986), the second sentence in (1a) introduces a
2
In the case of assertions, the topic time is the (implicitly or explicitly given) time
for which it is claimed that some state of affairs holds. If the utterance concerns
an event, the topic time may be different from the time of the event; the ordering
relation between topic time and event time (or “situation time”, in Klein’s terminology) is expressed by viewpoint-aspectual operators. For example, imperfective viewpoints place the topic time inside the time of the event, while perfective
viewpoints inversely place the time of the event inside the topic time. Klein’s
“topic time” corresponds broadly to the adaptation of Reichenbach’s (1947)
“reference point” in the DRT literature. In contrast, in the present framework,
“reference point” is used for the time interval with respect to which topic times
may be determined. A pure anaphoric tense on this account is an operator that expresses a relation between topic time and some reference point in the same way
a pure deictic tense expresses a relation between topic time and coding time.
Temporal anaphora in a tenseless language 3
new topic time (in the present terminology) following that of the first sentence, with the event time of the phone call included in the new topic time,
whereas the progressive in (1b) introduces a state to the discourse representation whose run time includes the topic time, and that reference time is unchanged from the first sentence, as if the second sentence tracked it anaphorically.3
The account sketched in this chapter presents the determination of topic
time vis-à-vis coding time in Yucatec as a special case of temporal anaphora.
The existence of temporal anaphora in a tenseless language such as Yucatec
is itself not surprising, but nevertheless remarkable: it shows that temporal
anaphora it is not an anaphoric meaning component of tense morphemes, as
assumed in Partee (1973) and the DRT literature. Topic times play a role in
the interpretation of utterances whether or not these are tensed, and the
principles involved in their contextual resolution are the same in tensed and
tenseless languages.
The account of temporal anaphora developed here treats the inferences
involved in topic time resolution as Gricean generalized conversational implicatures. Part of the evidence in favor of this approach is the non-monotonicity of the inferences. Contra Bittner (2008), the inferences are just as
defeasible in the “aspectually fully explicit” Yucatec as they are in English.
The Gricean analysis accounts for both the non-monotonicity of the inferences, which is attributed to vagueness in the DRT literature, and for their
default character. The Gricean account offers a parsimonious alternative to
the assumption of special principles governing the deictic interpretation of
tenseless sentences, as proposed in Smith, Perkins & Fernald (2007).
The following section presents a thumbnail sketch of the resources involved in temporal reference in Yucatec. Section 3 summarizes the evidence
for tenselessness, and section 4 lays out the analysis of temporal anaphora.
2.
A sketch of the Yucatec grammar and lexicon of temporal reference
2.1. Background
Yucatec is the largest member of the Yucatecan branch of the Mayan language family. It is spoken by 759,000 people in the Mexican states of Campeche, Quintana Roo, and Yucatán, and approximately 5,000 people in the
3
Hence the term “temporal anaphora” originally coined in Partee (1973) and identified as a metaphor in Partee (1984).
4 Jürgen Bohnemeyer
Cayo District of Belize (Ethnologue 2005).4 Yucatec is the language whose
autodenomination, Maya, has been adapted by scholars to name the Mayan
language family.
Yucatec is a polysynthetic language in the sense that syntactic relations
tend to have morphological reflexes at the word level and in the sense that
a single content word, in combination with the necessary function words
and inflections, may – and frequently does – constitute a clause. In terms of
the morphological complexity of content words, Yucatec is situated towards
the high end among Mayan languages, but in a more central position among
Mesoamerican languages overall. Yucatec is mostly head-initial, and in particular verb-initial, but this fact is somewhat obscured by the high frequency
of left-dislocations and focus constructions in discourse. The language has
a typologically unusual argument marking system in which the single core
argument of an intransitive clause patterns with either the actor or the undergoer argument of an active transitive clause depending on aspect-mood
inflection.5
There are five sources of overtly expressed temporal information in Yucatec: lexical items, the verb inflection system, adverbials, connectives, and
certain nominal suffixes. The following subsections address these in turn.
2.2. Lexical items
Yucatec has a complex system of lexical categories not all niceties of which
are as yet fully understood. The following thumbnail sketch summarizes
the analysis presented in Bohnemeyer (2002, 2004) and Bohnemeyer &
Brown 2007. This analysis proposes a taxonomic organization with a toplevel split between verbs and other categories. Verbs obligatorily inflect for
a functional category called status (following Kaufman 1990) in all syntactic environments in which they function as verbs. Status inflection expresses,
in a single suffix position, distinctions of viewpoint aspect, modality, and
4
5
Based on 2005 Census data, which register a decline by more than 40,000
speakers age five or older since 2000 (http://www.inegi.gob.mx/est/contenidos/
espanol/rutinas/ept.asp?t=mlen10&c=3337).
This is an atypical “split-S” or “active-inactive” pattern in which the morphological treatment of the core argument of intransitive clauses is fully determined,
not by the lexical semantic properties of the verb, but instead by inflectional distinctions akin to those found in split-ergative case marking systems; cf. Bohnemeyer 2004 and references therein.
Temporal anaphora in a tenseless language 5
illocution; cf. below. The architecture of the status system differs from language to language; in Yucatec, there are five subcategories: completive, incompletive, subjunctive, extra-focal, and imperative status. Every verb form
must be morphologically specified for exactly one of these subcategories in
any syntactic environment – there is no finiteness contrast with regard to
status inflection. In contrast, stative predicates are incompatible with status
inflection. All content words – other than verb roots or stems – can form
stative predicates by themselves, without the need of a copula.
(2)
Túumben le=nah=o’
new(B3SG) DET=house=D2
‘The house is/was/will be new’
Stative predicates include nouns, adjectives, and deverbal statives. Verb
stems fall into five different “conjugation” classes distinguished by paradigms of allomorphs of the status suffixes, which are termed “active”, “inactive”, “inchoative”, “positional”, and “transitive” in Bohnemeyer (2002,
2004).6 As shown in Bohnemeyer (2001, 2002, 2004), membership in these
verb stem classes is strikingly well motivated in terms of causativity and
the distinction between processes and state changes. Both processes and
state changes are dynamic eventualities that involve change in some property over time. Process descriptions either leave the event participant to
whom the changing property is attributed unexpressed (e.g., paint, as opposed to paint a picture/the wall) or fail to specify a reference point with
respect to which the change could be evaluated (e.g., walk, as opposed to
walk to the station); cf. Bohnemeyer & Swift 2006. The Yucatec active class
includes underived process verb stems – especially manner of motion (e.g.,
áalkab ‘to run’; bàab ‘to swim’) and sound emission verbs (e.g., hàayab ‘to
yawn; òok’ol ‘to cry’), denominal verb stems (e.g., e’l ‘egg’, ‘testicle’ > ‘to
ovulate’, ‘to lay eggs’; míis ‘broom’ > ‘to sweep’), antipassive stems derived from transitive roots (e.g., k’ay ‘to sing sth. (song)’ > k’àay ‘to sing’;
tus ‘to lie to sb.’ > tùus ‘to lie’), and all intransitive stems derived from
Spanish roots, regardless of their semantics – the most important exception
to motivation in terms of process semantics (e.g., áatrasáar ‘to become
late’; duràar ‘to last’; gáanar ‘to win’). The majority of process stems describe “internally caused” activities (including all of the examples cited
above); but externally caused events are likewise lexicalized in active roots
(though without encoding their cause; e.g., balak’ ‘roll’, chíik ‘shake’, ‘rattle’;
6
V denotes a vowel whose quality is copied from the preceding root vowel.
6 Jürgen Bohnemeyer
péek ‘move’, ‘wiggle’) (cf. Smith 1978 and Levin & Rappaport-Hovav 1995
on the distinction between internal and external causation).
Table 1. Status polysemy and verb stem classes
STATUS
STEM CLASS
INCOMPLETIVE
COMPLETIVE
SUBJUNCTIVE
EXTRAFOCAL
IMPERATIVE
active
-Ø
-nah
-nak
-nahik
-nen
inactive
-Vl
-Ø
-Vk
-ik
-en
inchoative
-tal
-chah
-chahak
-chahik
N/A
positional
-tal
-lah
-l(ah)ak
-lahik
-len
transitive
active
-ik
-ah
-Ø / -eh
-ahil
-Ø / -eh
passive
\’/ …-Vl
/ -a’l
\’/ …-ab
/ -a’b
\’/ …-Vk
/ -a’k
\’/ …-ik
/ -a’bik
N/A
Inactive, inchoative, and positional stems describe different kinds of state
changes without expressing their causes. Inchoative stems are derived from
adjectives and, less regularly, nouns. Positionals – or “dispositionals”
(Bohnemeyer & Brown 2007) – lexicalize non-inherent spatial properties
that may be thought of as “manners” of location. Distinctions that enter the
conceptualization of dispositions include “position”, or more generally
support/suspension (e.g., kul ‘sit’, wa’l ‘stand’, chil ‘lie’, xol ‘kneel’, nak
‘lean’, choh ‘hang’, hoch’ ‘droop’); blockage of motion (e.g., tak’ ‘be stuck
to something’, kap ‘be stuck between two things’); orientation in the gravitational field (e.g., haw ‘lie face up’, nok ‘lie face down’, tsel ‘lie on side’,
ch’eb ‘be tilted’, ‘lean to one side’); and configurations of parts of an object with respect to each other (e.g., hen ‘be sprawled’, nik ‘be scattered’,
hay ‘be spread out’, much’ ‘be in a pile’, tsol ‘be lined up in a row’, sop’
‘be coiled up’). Dispositional roots require derivational morphology for the
formation of both stative and intransitive dynamic stems, the latter expressing the “assume position” sense (Levin 1993). However, a majority of dispositional roots produces transitive verb stems without overt derivation.
Finally, inactive roots include “path verb” roots (bin ‘go’, tàal ‘come’, u’l
‘return’, máan ‘pass’, etc.); “phase” (= aspectual) roots (e.g., ho’p’ ‘begin’,
ts’o’k ‘end’); and also a few roots referring to bodily or cognitive state
changes (e.g., síih ‘be born’, kim ‘die’, ah ‘wake up’, wèen ‘fall asleep’)
and creation/destruction type events. The inactive class is fed by the anti-
Temporal anaphora in a tenseless language 7
causative derivation, and passivized transitive stems have a status pattern
closely related to that of inactive stems (see Table 1).7
Transitive roots overwhelmingly express caused state changes (e.g., kach
‘break’, xot ‘cut’), but a few also lexicalize forced contact (e.g. hats’ ‘hit’,
koh ‘beat’, yet’ ‘massage’) and changes in mental states in which an “experiencer” outranks the theme or stimulus (e.g., il ‘see’, na’t ‘guess’, ‘intuit’,
‘reason’, ‘understand’, and kan ‘learn’). All intransitive verb roots produce
derived transitive stems via the causative and “applicative” derivations.
There are very few, if any,8 stative verbs; meanings lexicalized in stative
verbs in English, such as ‘know’, ‘love’, or ‘have’, are expressed by relational nouns and elements of other categories. Thus, the parts-of-speech
system of Yucatec seems to be semantically motivated in terms of dynamicity more clearly than Indo-European languages.
Membership in the active class as opposed to the other verb stem classes
is determined along the lines of the process-change distinction, not in terms
of telicity, along the lines of the distinction between Vendlerian activities
vs. accomplishments and achievements (as suggested by Lucy 1994). Thus,
the inactive and inchoative verb stem classes include many “degree
achievements” projecting atelic event descriptions unless some “degree of
change” is overtly specified (Abusch 1985; Bertinetto & Squartini 1995;
Dowty 1979: 88-91; Hay, Kennedy, & Levin 1999). This is illustrated in (3)
for the inchoative kàabal-tal ‘lower’ (derived from the adjective kàabal
‘low’):
(3)
Le=mùunyal=o’ ts’u=chúun-ul
u=kàabal-tal,
DET=cloud=D2 TERM:A3=start\ACAUS-INC A3=low-INCH .INC
‘The cloud, it had started going down (lit. lowering),’
káa=h-tàal
le=ìik’=o’,
CON=PRV-come(B3SG) DET=wind=D2
‘(when/and then) the wind came’
káa=t-u=pul-ah
le=mùunyal=o’.
CON=PRV-A3=throw-CMP (B3SG) DET=could=D2
‘(when/and then) it drove (lit. threw) the cloude away.’
7
8
The anticausative is treated as a derivation in Bohnemeyer 2004, but is argued
to have also voice-like traits in Bohnemeyer 2007.
It is not clear that there are any verbs in the language that do not have at least
one dynamic sense.
8 Jürgen Bohnemeyer
Ts’u=kàabal-tal
le=mùunyal=o’?
TERM:A3=low-INCH .INC DET=cloud=D2
‘Had the cloud gone down?’
-Ts’o’k ka’ch
u=kàabal-tal
káa=h-pu’l-ih.
TERM formerly A3=low-INCH.INC CON=PRV-throw\PASS-CMB(B3SG)
‘It had gone down (when/and then) it was driven (lit. thrown) away.’
In (3) the speaker proposes a scenario to a native speaker consultant in
which a cloud starts sinking before it is pushed away by the wind. He asks
the consultant whether it is possible in this context to say that the cloud had
already become lower when it was pushed away. The consultant answers in
the affirmative.
Conversely, many active intransitive stems have salient telic “performance object” interpretations – especially antipassive stems such as k’àay
‘sing’ in (4):
(4)
Pedro=e’ táan u=k’àay,
Pedro=TOP PROG A3=sing\ATP
‘Pedro, he was singing,’
káa=t-u=k’at-ah
u=báah Pablo.
CON=PRV-A3=cross-CMP(B3SG) A3=self Pablo
‘(when/and then) Pablo interfered.’
Pedro=e’ t-u=p’at-ah
u=k’àay.
Pedro=TOP PRV-A3=leave-CMP(B3SG) A3=sing\ATP
‘Pedro’, he stopped singing.’
Be’òora=a’ ts’o’k=wáah u=k’àay
Pedro?
now=D2
TERM=ALT A3=sing\ATP Pedro
‘Now, has Pedro sung?’
- Ma’=h=bèey-chah
u=k’àay=i’.
NEG=PRV =thus-PROC.CMP(B .3.SG) A3=sing\ATP=D4
‘He didn’t manage to sing (lit. his singing didn’t become possible).’
In this case, the question asked of the consultant is whether somebody who
is singing, but interrupted in the course of it, can be said to have sung. The
consultant denies this, assuming that the singer in the scenario wasn’t able
to complete his song.
(3)–(4) illustrate the application of the entailment pattern known as the
“imperfective paradox” (Dowty 1979: 133) in telicity tests. There are no syntactic correlates of telicity in Yucatec. There is no distinction between dura-
Temporal anaphora in a tenseless language 9
tion (or for-type) and time-frame (or in-type) adverbials (cf., e.g., Krifka
1992); the same verbs are used in translations of “She spent X-time VERBing” and “It took her X-time to VERB”; and there are no “aspectualizers” or
“phase verbs” that only occur with either telic or atelic verbal cores (cf.
Freed 1979).9 It is not clear, at present, why telicity is reflected in the syntax of some languages but not in that of others.
2.3. The verb inflection system
In main clauses, the verb inflection system specifies two positions in which
aspectual and modal information is obligatorily expressed: status inflection
and the pre-verbal aspect-mood (AM) markers. Table 2 summarizes the
distribution and semantics of the five status categories.
Table 2. Distribution and semantics of the five status categories
Category
Distribution
Semantics
completive
independent verbal cores
w/ perfective AM marker
perfective aspect and
“assertive” modality
incompletive
dependent verbal cores;
independent verbal cores
w/ imperfective AM marker
imperfective aspect and
“assertive” modality
subjunctive
dependent verbal cores;
optative clauses; irrealis
subordinate clauses
perfective aspect and
“non-assertive” modality
extra-focal
manner focus construction
(dependent verbal core)
perfective aspect and
“assertive” modality
imperative
imperative sentences
perfective aspect and
“directive” illocution
Affirmative declarative main clauses are constituted by the combination of
a verbal core with exactly one of the 15 AM markers. Two of these – the
9
A verbal core is a maximal projection of a verb that dominates all of the verb’s
arguments and obliques, like a subject-internal verb phrase. The term is borrowed
from Van Valin & LaPolla 1997. There is no evidence of a (subject-external) verb
phrase in Yucatec. Verbal cores combine with aspect-mood markers to constitute
clauses in Yucatec (see below); clauses and verbal cores are thus in complementary distribution.
10 Jürgen Bohnemeyer
perfective and imperfective AM – are prefixes; the verbal cores they combine with are independent. The remaining 13 AM markers are stative predicates which take a dependent verbal core as their sole argument.10 Each
AM marker assigns a status category marked on the verb; Table 3 below
lists the combinations.11
The semantics of the preverbal AM markers is briefly discussed below
and in section 3. Relative and “topic” clauses (see section 2.5), focus constructions,12 and negation involve alternative AM systems. Details vary from
construction to construction (cf. Bohnemeyer 2002: 116–129); but all of
these constructions are governed by what may be characterized as a realisirrealis mood contrast, where realis mood includes reference to individual
present or past events, the irrealis future time reference, and habitual and
generic reference may be subsumed under either category depending on the
construction. In the realis mood, depending on the construction, either the
system in Table 3 or a simplified version involving bare incompletive and
subjunctive forms without preverbal AM markers are used. The irrealis is
expressed using a bare incompletive form under negation and the irrealis
subordinator kéen plus subjunctive status (in some cases also incompletive,
but only with intransitive verbs) in the other constructions. The examples in
(5) illustrate the realis-irrealis contrast for content questions targeting the
undergoer of a transitive verb. The realis example (5a) has the perfective
10
11
12
There is one exception: the prospective AM marker mukah. Mukah takes an
oblique dependent verbal core and carries a set-B suffix cross-referencing the
actor argument if the latter is transitive and the single argument if it is intransitive.
Status marking on the verb likewise depends on the verb’s transitivity (see Table
2), as it does in certain control constructions with lexical matrix predicates (see
Bohnemeyer in press).
Table 3 provides only the base forms of the AM markers; they frequently form
portmanteaus with the set-A clitics of the verbal core. Necessitive k’a’náan has
an apparent synonym k’abéet; some speakers use the two interchangeably, while
others prefer k’a’náan. Other, apparently more obsolete, general necessity modals
occur as well. The distinction between AM markers and lexical stative predicate
that combine with dependent verbal cores is not sharply delimited. For instance,
Ayres & Pfeiler (1997) and Vapnarsky (1999) treat sùuk ‘custom’, ‘habit’ as a
habitual marker, while Bohnemeyer (2002) considers sùuk a lexical stative complement-taking predicate.
There is evidence suggesting that all focus constructions, including all content
questions, are clefts in Yucatec and that all clefts use relative clause constructions as nominal predicates; cf. Bricker (1979); Bohnemeyer (2002: 116–129);
Tonhauser (2003). For an alternative viewpoint, cf. Gutiérrez Bravo (2006).
Temporal anaphora in a tenseless language 11
aspect marker with past time reference; the irrealis example (5b) the irrealis
subordinator kéen and subjunctive status on the verb with future time reference:
(5)
a. Ba’x
t-a=mèet-ah?
what(B3SG) PRV-A2=do:APP-CMP(B3SG)
‘What did you do?’, ‘What is it that you did?’
b. Ba’x
kéen a=mèet-eh?
what(B3SG) SR.IRR A2=do:APP-SUBJ(B3SG)
‘What will you do/are you going to do?’, ‘What is it that you will
do/are going to do?’
Table 3. Preverbal AM markers and status assignment
Subset of AM markers
AM marker
Status category
AM prefixes
perfective h-(Vitr)/t- (Vtr)
imperfective k-
completive
incompletive
Aspectual AM
predicates
progressive táan
terminative ts’o’k
prospective mukah
Modal AM
predicates
Metrical AM
predicates
obligative yan
necessitive k’a’náan
desiderative táak
assurative he’ …=e’
penative òolak
incompletive (Vitr) /
subjunctive (Vtr)
incompletive
subjunctive
remote future bíin
remote past úuch
recent past sáam
immediate past táantik …=e’ incompletive
proximate future ta’itak
Finally, dependent verbal cores also occur as arguments or obliques of lexical predicates and as complements of prepositions. In these constructions,
status marking is controlled by the head (cf. Bohnemeyer 2002: 91–101 for
details).
It is conceivable that the presence vs. absence of an AM marker in a
clause or phrase is correlated with – and thus an expression of – the semantic property of finiteness. Finiteness can be understood as the property of a
clause or verbal projection of requiring a particular “topic component” for
12 Jürgen Bohnemeyer
its interpretation (Klein 2006). In languages that distinguish finite from nonfinite verb forms, projections of finite verb forms must be understood as
making an assertion, asking a question, etc., about a particular topic time, a
particular topic place, (a) particular possible world(s), a particular set of
entities, etc. In contrast, projections of nonfinite verb forms are interpreted,
not with respect to their own topic components, but as part of some superordinate syntactic structure which is ultimately “hooked up” to a topic
component via a finite verb form. Yucatec lacks a distinction between finite and non-finite verb forms. All verb forms are marked for status, and all
status categories except for the completive occur both in independent and
dependent verbal cores. It seems plausible that, in first approximation, (dependent or independent) verbal cores are inherently non-finite and that it is
the combination with a preverbal AM marker that maps the “sentence base”
expressed by the verbal core into a topic component. Future research will
have to examine how finiteness is expressed – if at all – in focus constructions, etc., and under negation.
The semantics of the five status categories is analyzed in Bohnemeyer
(2002: 216–242), with the upshot represented in Table 2. The perfectiveimperfective distinction is conceptualized here following Klein 1994, i.e., in
terms of a temporal relation between the time of the eventuality described
in an utterance and the “topic time” for which the denotation of the utterance is computed. Perfective aspect places the event time inside the topic
time; imperfective aspect conversely places the topic time inside the event
time. Subjunctive status, like completive, is perfective, i.e., encodes inclusion of the event time of the verbal core in the topic time. The subjunctive
differs from the completive in that it “decouples” the interpretation of the
sentence or clause formed by combining the verbal core with an AM marker
or a lexical matrix predicate from the realization of the event described by
the verbal core (“non-assertive” modality). The subjunctive occurs as an
optative mood with matrix predicates of desire. It is likewise triggered by
the “penative” AM marker (from Latin paene “almost”), which has a counterfactual semantics and is used to describe events that almost, but not quite,
become reality. The remote future and recent and remote past markers,
which likewise govern the subjunctive, cardinally quantify over the distance between topic time and event time. None of these entails event realization (cf. Bohnemeyer & Swift 2004). The immediate, recent, and remote
past markers presuppose, rather than to assert, realization of the event described by the verbal core. Hence, when they are negated, it is to deny the
recency/remoteness of the event described by the verbal core – not that the
event happened:
Temporal anaphora in a tenseless language 13
(6)
Ma’ sáam sùunak
le=kòombi=o’;…
NEG REC turn\ATP:SUBJ(B3SG) DET=van=D2
‘It’s not a while ago that the bus returned;…’
a. …inw=a’l-ik=e’,
h-ts’o’k
mèedya òora.
A1SG=say-INC(B3SG)=TOP PRV-end(B3SG) half
hour
‘…I think it was half an hour ago.’13
b. ??…tuméen ma’ sùunak=i’.
CAUSE
NEG turn\ATP:SUBJ(B3SG)=D4
‘…because it hasn’t returned yet.’
Presupposition may also motivate the occurrence of the bare subjunctive in
certain focus constructions (including the so-called “agent-focus” form)
with perfective reference. The irrealis subordinator kéen illustrated above
likewise assigns subjunctive status.
“Extra-focal” is a status category exclusively reserved to a particular kind
of focus constructions: manner focus constructions, in which the focused
element refers to a property of the event described by the “extra-focal”
subordinate clause. If that clause is perfective, its verb is in the extra-focal.
Example (7) illustrates:
(7)
Domìingo-ak=e’ ma’+lo’b h-hàats’-nahik-en.
Sunday-CAL =TOP NEG+bad PRV-hit\ATP-EXTRAFOC-B1SG
‘Last Sunday, well is how I batted.’
Imperative forms are used exclusively in commands. They do not combine
with preverbal AM markers, and the set-A prefixes cross-referencing the
addressee as the single argument of intransitive imperatives and the A-argument of transitive imperatives is deleted.
Turning to the AM markers, the semantics of the five aspectual AM
markers is sketched in Figure 1. The analysis presupposes the framework of
Klein 1994, with one modification: prospective and perfect are understood
as placing topic time in the run time of some pre- or post-state causally related to the target event, rather than merely in times preceding and following the event, respectively. The imperfective AM marker has progressive
13
The recent past marker is a “same day” past marker, i.e., is used for events that
happened earlier on the day that includes topic time. By preemption, it receives
a non-immediacy implicature due to its contrast with the immediate past marker.
This implicature is interpreted in (6b) as metalinguistically negated.
14 Jürgen Bohnemeyer
interpretations mostly in focus constructions. Outside focus constructions, it
occurs in irrealis contexts, i.e., in clauses with habitual, generic, and future
time reference whose subordinate counterparts trigger irrealis marking in
the appropriate syntactic environments. Only the progressive reading is
represented in Figure 1. The perfective AM marker has both perfective interpretations proper and (resultative) perfect interpretations; Bohnemeyer
(2002: 246–250) argues that this is a case of vagueness. The question
whether the perfective marker incorporates a past tense component is discussed in section 3.
Four of the five modal AM markers involve universal quantification over
the possible worlds in “modal bases” (Kratzer 1977, 1981, 1991) of different
kinds. The obligative is used to express that the sentence base eventuality is
realized in all worlds consistent with the social obligations of the referent
of the S/A argument, with schedules, and with the ordinary course of events
in general. The desiderative AM marker indicates that the base eventuality
is realized in all possible worlds in which the (in particular, bodily) desires
of the referent of the S/A argument are met. The assurative AM is used for
universal quantification over a modal base in which promises, offers, and
commitments by the referent of the S/A argument are fulfilled; in these uses,
it conveys a certain illocutionary flavor. But the assurative is also used with
epistemic modal bases. In this case, it expresses certainty; but, perhaps due
to the contrast with the obligative and prospective markers – both of which
are used for reference to events expected to occur in “inertia worlds”
(Dowty 1979; Landman 1992; Portner 1998), based on evidence of causally
related pre-states that obtain in the topic situation – epistemic uses of the
assurative tend to have a strong flavor of subjective opinion or estimate.
The fourth universal-force modal, the necessitive, is used to express circumstantial or conditional necessity – the idea that something must happen if
something else is to be allowed to happen or to be avoided.14
14
While there are thus four universal-force modals integrated into the paradigm of
AM markers, existential modality – i.e., possibility – is expressed by inflected
lexical predicates. The reason for this curious asymmetry is unknown.
Temporal anaphora in a tenseless language 15
pre-state sPRE
CAUSE (sPRE,e)
run time τ (sPRE)
of the pre-state
result state sPOST
CAUSE (e,sPOST)
eventuality e
event time τ(e)
prospective AM:
tTOP ⊂ τ (sPRE)
progressive and imperfective AM: tTOP ⊂ τ (e)
run time τ (sPOST)
of the post-state
time t
terminative and
perfective AM:
tTOP ⊂ τ (sPOST)
perfective AM:
τ (e) ⊆ tTOP
Figure 1. Semantics of the aspectual AM markers (key: CAUSE(a,b) – event a
is/was/will be the/a cause of event b; SPRE – pre-state; SPOST – result-state;
τ(a) – run time of a; tTOP – topic time)
The “penative” AM marker òolak is used to indicate that an instance of the
base event was at some past topic time about to happen, but subsequently
did not – a purely counterfactual sense of “almost”. It can be viewed as a
counterfactual mirror image of the prospective and proximate-future markers. Its grouping with the four universal-force modals is obviously somewhat arbitrary; it does not perfectly fit into any of the subcategories of AM
markers distinguished here.
The five metrical AM predicates quantify the distance between the run
time of the base eventuality and topic time. The denotation of proximity in
the proximate future marker (how soon is “soon”?) is purely a matter of
contextual standards. The rare remote future marker seems to be primarily
used to convey negation of this subjective sense of proximity. Consultants
consistently volunteer the information that the remote future suggests that
the speaker has no idea when, if at all, the base event will happen.15 The
15
The remote future marker was mischaracterized as a “predictive” modal marker
in Bohnemeyer (1998b, 2000a, 2002, 2003). The research on which this body of
work was based did not produce evidence of the marker. The conjecture of modal
semantics took off from the observation that the marker is associated with the
genre of prophecies (Vapnarsky 1995). However, recent elicitation has shown it
to be part of the “degrees-of-remoteness” system. See section 3 for examples.
16 Jürgen Bohnemeyer
three metrical past markers form a scale16 the central member of which, the
recent past marker sáam, prototypically extends into the relative past up to
the boundary between the day that includes topic time and the preceding
day, i.e., functions as a “hodiernal past” marker.17 By preemption, the immediate and remote past markers are interpreted to the exclusion of the recent past marker. In the case of the immediate past marker táantik …=e’,
this has the effect that the latter is used predominantly for events that happened at a distance which is short compared to the lengths of a day (no
more than a few hours, and often less than an hour), while the remote past
marker úuch is used at a distance considered long by the same standard
(typically several days ago).
The five metrical distance predicates differ in several ways from “metrical tense” phenomena described in the literature (e.g., Comrie 1985: ch.4;
Dahl 1984, 1985: 120 –128). It is argued in section 3 that instead of constraining possible topic times of the utterance vis-à-vis coding time or some
reference point, as true (metrical) tenses would, they express the distance
between topic time and event time as a state that obtains at topic time.
2.4. Adverbials
Within the domain of temporality, adverbials are used to express calendrical
or clock time indices or the event times of eventualities. There is a handful
of lexical indexical adverbs, including terms locating days with respect to
coding time (e.g., ka’ho’lheak ‘the day before yesterday’, sáamal ‘tomorrow’) and the “topic time shifters” be’òora …=a’/=e’ ‘now’ and ka’ach(il)
‘formerly’. The latter are briefly discussed in section 3. Apart from these,
temporal adverbials are mostly headed by spatial prepositions such as ich(il)
‘in’, ‘inside’ and relational nouns such as táan ‘front’. These adverbials
specify time intervals, but do not encode temporal ordering relations between these times and the topic or event time of the utterance. Thus, ich(il
ti’) ts’e’ets’ek k’ìin may be used to translate ‘in a few days’ (referring to an
16
17
This is not an entailment scale in the sense of Horn 1972 since the three terms
denote distances from topic time and thus entail upper bounds. The markers’ interpretation is nonetheless subject to pragmatic enrichment, as mentioned below.
This cut-off point is blurred by the frequent use of sáam to express that something happened earlier than expected (by the speaker and/or the addressee), even
if it happened within the domain of the immediate or remote past. See also section 3.
Temporal anaphora in a tenseless language 17
event in the (relative) future), ‘((with)in a few days’ (referring to a time
span within which some event was or will be completed), ‘for a few days’
(referring to the duration of a process or interval), or ‘a few days ago’, depending on aspect-mood marking.
(8)
Pwes to’n =e’,
ich ts’e’ts’ek k’ìin hóok’-ok-o’n.
well PREP:B1PL=TOP in a.few
sun exit-SUBJ-B1PL
‘Well, as for us, it was a few days ago that we left.’
(9)
Pwes to’n =e’,
ich ts’e’ts’ek k’ìin kéen hóok’-ok-o’n.
well PREP:B1PL=TOP in a.few
sun SR.IRR exit-SUBJ-B1PL
‘Well, as for us, it is in a few days that we will leave.’
Bohnemeyer (in press) suggests that this may be a reflex of the literal spatial
meanings of the heads, which encode what Jackendoff 1983 calls “place
functions”, but no locative or path relations. For instance, ich le=nah=o’,
formed with the same preposition ich, translates ‘in the house’, ‘into the
house’, ‘out of the house’, and ‘through the house’, depending on the verb
the phrase combines with.
2.5. Connectives
As in other languages, subordinate clauses may serve to introduce the time
of the event they describe as a point of reference in discourse. There are,
however, no connectives in Yucatec that encode ordering relations between
time intervals such as ‘after’, ‘before’, or ‘while’. There is likewise no
word for ‘when’, although there are a few idioms that serve as temporal
anaphors without expressing ordering relations. An example is the phrase
chéen ya’lo’ ‘at that time’, literally ‘it will say that’ in (10), using a form of
the irrealis subordinator kéen:
(10) Chéen ka’=sùunak-ech
t-u=láak’
ha’b=e’,
SR:IRR REP=turn\ATP:SUBJ-B2SG PREP-A3=other year=TOP
chéen y=a’l-∅=e’,
SR:IRR A3=say-SUBJ(B3SG)=TOP
táantik in=mèet-ik
le=nah=o’.
IMM
A1SG=DO:APP-INC (B3SG) DET=house=D2
‘(When) you return here next year, at that time (lit. (when) it will say
that), I will have just build the house.’
18 Jürgen Bohnemeyer
The first clause in (10), chéen ka’ sùunakech tuláak’ ha’be’ ‘(when) you
return next year’, introduces a reference point taken up first by the phrase
that functions as a temporal anaphor (itself a clause) – which is actually
redundant in this example – and then by the main clause, whose topic time
it is understood to mark. These subordinate clauses determine the topic
time for the main clause much like temporal clauses in Indo-European languages. However, syntactically, they are not adverbials, but share the morphological and ordering properties of left-dislocated arguments – they are
“topic clauses” (Bohnemeyer 1999c). Neither the clause that introduces the
reference time nor the phrase chéen ya’lo’ are morphologically “flagged”
for expressing temporal reference points. They are marked as topics by
their position and the clause-final particle that follows them (the topic
marker =e’ in (10)); but in an appropriate context, the same clauses could
be interpreted as headless relative clauses. This ambiguity is illustrated in
(11):
(11) Le=k-u=tàas-a’l=e’,
k-u=bo’l-t-a’l.
DET=IMPF-A3=come:CAUS-PASS.INC=TOP IMPF-A3=pay-APP-PASS.INC
‘(What) is brought is paid for / (when) it is brought, it is paid for.’
(based on Blair & Vermont Salas 1965–67 11.1.25)
Instead of connectives expressing ordering relations, both “topic clauses”
of the kind illustrated in (10)–(11) and main clauses frequently carry what
may be described as “aspectual connectives” (Bohnemeyer 1998b: 485 –
503). An example is káa, which occurs exclusively with the perfective AM
marker. The perfective AM marker has both perfective and result-state interpretations (e.g., the same clause can be used to convey ‘The car broke down’
and ‘The car is broken down’, not unlike the simple past in English). Káa
forces the perfective interpretation. Result-state reference, like all notional
viewpoint aspects except for the perfective, requires a contextual reference
point (a “natural temporal reference point”, see section 4) to determine its
topic time. Káa blocks interpretation of the clause with respect to a tracked
topic time. Put differently, it signals “resetting” of the topic time variable.
This device is used above all in narrative discourse. The result can be a sequential or simultaneous interpretation depending on context. In (12), the
preferred interpretation is sequential. The order of events is inferred from
the order of clauses on the basis of stereotype implicatures in this case (see
section 4). In (13), the preferred interpretation is simultaneous for most
speakers because the two actions reported have different agents. This makes
it clear that káa does not express an ordering relation.
Temporal anaphora in a tenseless language 19
(12) Pedro=e’ káa=t-u=ts’íib-t-ah
hun-p’éel
Pedro=TOP CON=PRV-A3=write-APP-CMP (B3SG) one-CL.IN
kàarta=e’, káa=t-u=ts’u’uts’-ah
hun-p’éel chamal
letter=TOP CON=PRV-A3=suck-CMP (B3SG) one-CL.IN cigarette
‘Pedro, (when) he wrote a letter, he smoked a cigarette’
(preferred interpretation sequential)
(13) Káa=t-u=ts’íib-t-ah
hun-p’éel kàarta Pedro=e’,
CON=PRV-A3=write-APP-CMP (B3SG) one-CL.IN letter Pedro=TOP
Juán=e’, káa=t-u=ts’u’uts’-ah
hun-p’éel chamal
Juán=TOP CON=PRV-A3=suck-CMP (B3SG) one-CL.IN cigarette
‘(When) Pedro wrote a letter, Juán smoked a cigarette’
(preferred interpretation for most consultants simultaneous)
In (12)–(13), the topic clauses refer to the times of the events they describe.
The topic times of the main clauses are interpreted with respect to these
event times, though not necessarily as coinciding with them, since the main
clauses likewise contain the “topic time reset” connective káa. The results
are the observed interpretations. It is also possible for topic clauses to refer
to times before or after the event they describe. Aspectual and modal expressions are used for this purpose. Example (14) shows a topic clause that
sets a topic time for the main clause preceding the topic clause event, the
speaker’s arrival:
(14) Ma’ k’uch-uk-en=e’,
káa=h-hóok’
leti’
NEG arrive-SUBJ-B1SG =TOP CON=PRV-exit(B3SG) DET:PREP
‘(When) I had not yet arrived, (and) she left’
The topic clause in this case uses negation and subjunctive status to refer to
a time prior to the realization of the topic clause event. This construction is
the closest Yucatec equivalent of the English connective before.
This section has provided an overview of the various types of lexical
and morphosyntactic resources involved in the expression of temporality in
Yucatec. The following sections summarize the evidence for tenselessness
(section 3) and offer an account of temporal anaphora in Yucatec (section 4).
20 Jürgen Bohnemeyer
3. Tenselessness
Yucatec is a tenseless language in the sense that the morphosyntactic form
of the clause does not constrain its use with topic times in the present, past,
or future of coding time (absolute = deictic tense) or some other reference
point (relative = anaphoric tense). “Morphosyntactic form” is taken here to
mean syntactic structure plus/including inflection. To demonstrate tenselessness, it needs to be shown that any Yucatec clause regardless of syntactic
structure and inflection is compatible with both topic times in the present,
past, and future of coding time and topic times in the present, past, and future of some other reference point. The discussion in the present chapter
focuses on clauses that can constitute sentences by themselves (“main” or
“independent” clauses). Subordinate clauses are briefly taken up at the end
of this section. The syntactic variation that needs to be taken into account
when evaluating the claim of tenselessness for main/independent clauses is
primarily the selection of the preverbal “aspect-mood” (AM) marker. As
mentioned in the previous section, the “status” suffix on the verb is determined by the AM marker. The examples in (15) illustrate the 15 AM markers.
(15) a. Perfective AM – completive status
T-in=mèet-ah
le=nah=o’
PRV-A1SG=do:APP-CMP(B3SG) DET=house=D2
‘I built the house’
b. Imperfective AM – incompletive status
K-in=mèet-ik
le=nah=o’
IMPF-A1SG =do:APP-INC(B3SG ) DET=house=D2
‘I (would) build the house’
c. Terminative AM – incompletive status
Ts’o’k in=mèet-ik
le=nah=o’
TERM A1SG=do:APP-INC(B3SG) DET=house=D2
‘I (will) have/had built the house’
d. Progressive AM – incompletive status
Táan in=mèet-ik
le=nah=o’
PROG A1SG=do:APP-INC(B3SG) DET=house=D2
‘I am/was/will be building the house’
Temporal anaphora in a tenseless language 21
e. Prospective AM – subjunctive status on transitive verbs, incompletive on intransitive verbs
Mukah in=mèet-∅
le=nah=o’
PROSP A1SG=do:APP-SUBJ(B3SG) DET=house=D2
‘I am/was/will be going to build the house’
f. Necessitive AM – incompletive status
K’a’náan in=mèet-ik
le=nah=o’
NEC
A1SG=do:APP-INC(B3SG) DET=house=D2
‘I must/had to/will have to build the house’
g. Obligative AM – incompletive status
Yan in=mèet-ik
le=nah=o’
OBL A1SG=do:APP-INC(B3SG) DET=house=D2
‘I (will) have/had to build the house’
h. Assurative AM – incompletive status
He’ in=mèet-ik
le=naho’
ASS A1SG=do:APP-INC(B3SG) DET=house=D2
‘I will/would indeed (agree to) build the house (you shall see)’
i. Desiderative AM – incompletive status
Táak in=mèet-ik
le=nah=o’
DES
A1SG=do:APP-INC(B3SG) DET=house=D2
‘I (will) want(ed) to build the house’
j. Penative AM – subjunctive status
Òolak in=mèet-∅
le=nah=o’
PEN
A1SG=do:APP-SUBJ(B3SG) DET=house=D2
‘I (will have/had) almost built the house’
k. Remote future AM – subjunctive status
Bíin in=mèet-∅
le=nah=o’
REMF A1SG=do:APP-SUBJ(B3SG) DET=house=D2
‘It is/was/will be a long time before I build the house’
l. Proximate future AM – incompletive status
Ta’itak in=mèet-ik
le=nah=o’
PROX
A1SG=do:APP-INC(B3SG) DET=house=D2
‘I will/would soon build the house’
m. Immediate past AM – incompletive status
Táantik in=mèet-ik
le=nah=o’
IMM
A1SG=do:APP-INC(B3SG) DET=house=D2
‘I (had/will have) just built the house’
22 Jürgen Bohnemeyer
n. Recent past AM – subjunctive status
Sáam in=mèet-∅
le=nah=o’
REC
A1SG=do:APP-SUBJ(B3SG) DET=house=D2
‘I (had/will have) built the house already/not long ago’
o. Remote past AM – subjunctive status
Úuch in=mèet-∅
le=nah=o’
REMP A1SG=do:APP-SUBJ(B3SG) DET=house=D2
‘I (had/will have) built the house long ago’
The 15 examples can be used, with the exceptions to be discussed below, in
each of the following three contexts in the position marked “_____”. Each
context determines the topic time of the utterance following it in the position
marked “_____”. In (16), Jorge is asking Pedro whether his has built the
house he had been planning to build. This question introduces a time frame
starting with Jorge’s last visit two years earlier and leading up to the present
of the utterance. In (17), Jorge asks whether Pedro will build the house in
the future, and Pedro begins his response by shifting the topic time to
Jorge’s next visit a year into the future. In the final context, (18), Jorge asks
whether Pedro’s house is new, and Pedro sets the topic time of his response
to the time of Jorge’s last visit two years earlier.
(16) Diagnostic context: topic time leading up to utterance time
Jorge’
táantik u=k’uch-ul
x-Yaxley.
Jorge:TOP IMM A3=arrive-INC F-Yaxley
‘Jorge, he has just arrived in (the village of) Yaxley.’
H-ts’o’k
ka’-péel ha’b káa=h-sùunah
PRV-end(B3SG) two-CL.IN year CON=PRV-turn:CMP (B3SG)
‘It has been two years since he returned…’
t-u=kàah-al=o’.
PREP-A3=reside\ATP-REL =D2
‘…to his country.’
T-uy=ohel-t-ah
PRV-A3=knowledge-APP-CMP (B3SG)
‘He knew…’
táak u=mèet-ik
u=nah-il
Pedro.
DES A3=do:APP-INC (B3SG) A3=house-REL Pedro
‘…that Pedro wanted to build a (lit. his) house.’
Temporal anaphora in a tenseless language 23
Ba’x=e’ ma’ t-uy=ohel-t-ah
what=TOP NEG PRV-A3=knowledge-APP-CMP(B3SG)
‘But he didn’t know…’
wáah h-bèey-chah-ih.
ALT PRV-thus-INCH.CMP-CMP(B3SG)
‘…whether he was able to do it (lit. whether it became possible).’
Káa=t-uy=il-ah
Pedro te=kàaye=o’,
CON=PRV-A3=see-CMP(B3SG) Pedro PREP:DET=street=D2
‘He saw Pedro in the street,’
káa=t-u=k’áat+chi’-t-ah.
CON=PRV-A3=wish+mouth-APP-CMP(B3SG)
‘and he asked him.’
Káa=t-u=núuk-ah
Pedro=e’:
CON=PRV-A3=answer-CMP (B3SG) Pedro=TOP
‘Pedro answered:’
“_____”
(17) Diagnostic context: topic time in the future of utterance time
Jorge’
ta’itak u=sùut
t-u=kàah-al=o’.
Jorge:TOP IMM A3=turn\ATP PREP-A3=reside\ATP-REL =D2
‘Jorge, he would soon return to his country.’
T-uy=ohel-t-ah
PRV-A3=knowledge-APP-CMP(B3SG)
‘He knew…’
táak u=mèet-ik
u=nah-il
Pedro.
DES A3=do:APP-INC(B3SG) A3=house-REL Pedro
‘…that Pedro wanted to build a (lit. his) house.’
Ba’x=e’ ma’ t-uy=ohel-t-ah
what=TOP NEG PRV-A3=knowledge-APP-CMP(B3SG)
‘But he didn’t know…’
wáah yan u=bèey-tal.
ALT OBL A3=thus-INCH .INC
‘…whether he would be able to do it (lit. whether it would become
possible).’
Káa=t-uy=il-ah
Pedro te=kàaye=o’,
CON=PRV-A3=see-CMP(B3SG) Pedro PREP:DET=street=D2
‘He saw Pedro in the street,’
24 Jürgen Bohnemeyer
káa=t-u=k’áat+chi’-t-ah.
CON=PRV-A3=wish+mouth-APP-CMP(B3SG)
‘and he asked him.’
Káa=t-u=núuk-ah
Pedro=e’:
CON=PRV-A3=answer-CMP(B3SG) Pedro=TOP
‘Pedro answered:’
“Chéen ka’=sùunak-ech
t-u=láak’
ha’b=e’,
SR:IRR REP=turn\ATP:SUBJ-B2SG PREP-A3=other year=TOP
‘When you return next (lit. the other) year,…’
_____”
(18) Diagnostic context: topic time in the past of utterance time
Jorge’,
t-uy=ohel-t-ah
Jorge=TOP PRV-A3=knowledge-APP-CMP(B3SG)
‘Jorge, he learned…’
t-u=mèet-ah
u=nah-il
Pedro.
PRV-A3=do:APP-CMP(B3SG) A3=house-REL Pedro
‘…that Pedro had built a (lit. his) house.’
káa=t-u=k’áat+chi’-t-ah
CON=PRV-A3=wish+mouth-APP-CMP(B3SG)
‘He asked him…’
wáah túumben le=nah=o’.
ALT new(B3SG) DET=house=D2
‘…whether the house was new.’
Káa=t-u=núuk-ah
Pedro=e’:
CON=PRV-A3=answer-CMP(B3SG) Pedro=TOP
‘Pedro answered:’
“Káa=h-tàal-ech
way h-ts’o’k
ka’=p’éel ha’b=e’,
CON=PRV-come-B2SG here PRV-end(B3SG) two=CL.IN year=D3
‘When you came here two years ago,…’
_____”
Most of the utterances in (15) are readily interpretable in these three contexts. For illustration, (19) has the remote past AM marker with future time
reference and (20) the remote future AM marker with past time reference:
Temporal anaphora in a tenseless language 25
(19) Chéen ka’=sùunak-ech
t-u=láak’
ha’b=e’,
SR:IRR REP=turn\ATP:SUBJ-B2SG PREP-A3=other year=TOP
úuch in=mèet-∅
le=nah=o’
REMP A1SG=do:APP-SUBJ(B3SG) DET=house=D2
‘When you return next (lit. the other) year, I will have built the house
long ago’
(20) Káa=h-tàal-ech
way h-ts’o’k
ka’=p’éel ha’b=e’,
CON=PRV-come-B2SG here PRV-end(B3SG) two=CL.IN year=D3
bíin in=mèet-∅
le=nah=o’
REMF A1SG=do:APP-SUBJ(B3SG) DET=house=D2
‘When you came here two years ago, it was going to be a long time
before I would build the house’
However, there are a number of principled exceptions. First of all, there are
pragmatic issues that may affect the naturalness of various markers in these
contexts.18 Secondly, the imperfective aspect marker (15b) has a progressive
meaning in focus constructions, but outside those is compatible with all and
only those topic times that in subordinate clauses trigger irrealis marking
(see section 2.3): habitual, generic, and future topic times. It is thus anomalous in descriptions of the building of an individual house in present (16)
and past contexts (18), but is interpretable in the future time reference context of (17). Finally, and most significantly for present purposes, the perfective (15a) and “penative” (counterfactual) AM markers (15j) are incompatible with future time reference. The most straightforward explanation of
this restriction would be that the markers have past (or non-future; the perfective is compatible with blow-by-blow-reporting-style contexts) meaning
components. However, the perfective AM marker is in fact used with future
time reference in conditional antecedents:
18
It is of course difficult to find a single example sentence format such that all 15
markers are equally natural in this format in all three contexts. The recent past
marker (15n) is considered infelicitous by some consultants in all three contexts
because it is commonly used to indicate that the target event happened/was realized earlier than expected/hoped/feared etc. This presupposition makes no sense
in the contexts in (16)–(18). Most of the modal markers are more likely to be
accepted with topic times shifted into the past or future of utterance time when
occurring in the complement of a verb of cognition or a speech act verb or the
like. The perfective AM marker is used for both perfective and result state reference. In the narrative context (18), the connective káa is preferred to be added at
the left edge of (15a) to force the perfective interpretation.
26 Jürgen Bohnemeyer
(21) a. Wáah t-a=ts’o’k-s-ah
le=nah
ALT
PRV-A2=end-CAUS-CMP (B3SG ) DET=house
te=mèes
k-u=tàal=o’,
PREP:DET=month IMPF-A3=come=D2
hi’n=bo’l-t-ik
tèech
be’òora=a’.
ASS:A1SG=pay-APP-INC (B3SG) PREP:B2SG now=D1
‘If you build the house next month, I will pay you now.’
b. Wáah káa=ts’o’k-s-∅
le=nah
ALT
SR:A2 =end-CAUS-SUBJ(B3SG) DET=house
te=mèes
k-u=tàal=o’,
PREP:DET=month IMPF-A3=come=D2
hi’n=bo’l-t-ik
tèech
be’òora=a’.
ASS:A1SG=pay-APP-INC(B3SG ) PREP:B2SG now=D1
‘If you build the house next month, I will pay you now.’
In (21), a hypothetical conditional is used to offer the addressee a deal. For
this purpose, either the perfective (a) or the subordinator káa in combination
with subjunctive status (b) can be used without discernible semantic difference. Notice that the perfective does not convey relative past time reference
in (21) either: the reference point for an anaphoric-tense interpretation
would have to be the event time of the main clause, which however lies in
the relative past, not future, of the topic time of the conditional clause in
(21). If the semantic contribution of the perfective in conditionals such as
(21a) is compositional, then the perfective cannot have a (deictic or anaphoric) past tense semantics.
The perfective AM marker occurs with future time reference in conditional antecedents, but not in any clause that asserts, questions, or presupposes propositions. I have argued (Bohnemeyer 1998b, 2000a, 2002) that
this is due to a combination of the fact that the perfective marker entails
event realization (Bohnemeyer & Swift 2004), i.e., factuality, with a principle that bars treating the realization of future events as fact:
(22) Modal commitment constraint (MCC)
The realization of events in the (relative or absolute) future cannot be
asserted, denied, questioned, or presupposed as fact. Assertions, questions, and presuppositions regarding the future realization of events, or
the failure thereof, require specification of a modal attitude on the part
of the speaker.
Temporal anaphora in a tenseless language 27
The MCC requires clauses used to assert, deny, question, or presuppose
realization of future events to express a modal attitude towards the realization of the event: necessity, desire, agreement, prediction, etc. Since the
modal markers that express these attitudes cannot co-occur with the perfective AM marker in the same clause, the perfective is excluded from the
relevant contexts by the MCC. The MCC is a language-specific constraint.
However, similar principles have been reported for other tenseless languages (Comrie 1985: 50 –53 for Burmese and Dyirbal; Bittner 2005 for
Kalaallisut /West Greenlandic).
The MCC accounts the unavailability of the “penative”, i.e., counterfactual, AM marker with future time reference since this marker negates
realization of the target event. But why does the MCC not exclude the terminative aspect marker and the immediate, recent, and remote past markers
from occurring with future time reference? All of these markers entail realization of the event described by the verbal core. However, these markers
have stative meanings: they serve to assert, deny, or question, not the realization of the target event, but the result state of the target event (terminative
AM) or the state of the target event’s immediacy (immediate past AM), recency (recent past AM), or remoteness (remote past AM). Stativity of these
markers is evident from their incompatibility with event time adverbials.
This is illustrated for the terminative AM marker in (23). The adverbial
ho’lheak ‘yesterday’ can only be interpreted as a topic time adverbial in
(23a), rendering the sentence both pragmatically (the speaker is asking
whether the addressee was during the day before the utterance in the state
of having met their brother) and syntactically odd (topic time adverbials are
preferred to be left-dislocated). To obtain the event time interpretation (i.e.,
to ask whether the addressee met the speaker’s brother the day before the
utterance), the perfective AM marker is used, as in (23b):
(23) a. ?Ts’o’k aw=il–ik
in=suku’n
ho’lheak?
TERM
A2=see-CMP (B3SG) A1SG=older.brother yesterday
‘Had you met my brother yesterday?’
b. T-aw=il-ah
in=suku’n
ho’lheak, he’bix
PRV-A2=see-CMP(B3SG) A1SG=older.brother yesterday like
t-a=tukul-ah=e’?
PRV-A2=think-CMP (B3SG)=D3
‘Did you meet my brother yesterday, as you had planned?’
Incompatibility with event time adverbials can be established in the same
fashion for the immediate, recent, and remote past AM markers. This incom-
28 Jürgen Bohnemeyer
patibility suggests that it is only the result state of the event in the case of
the terminative marker and the state of the event having occurred at a certain distance from topic time in the case of the metrical tense markers that
is accessible to adverbial modification. And it is the stativity of the markers
in question that explains why they are not barred from future topic times by
the MCC. All stative predicates of Yucatec occur freely with arbitrary topic
times – the MCC does not apply to state descriptions:
(24) (Káa=h-tàal-ech
way h-ts’o’k
ka’=p’éel ha’b=e’,
CON=PRV-come-B2SG here PRV-end(B3SG) two=CL.IN year=D3
/chéen ka’=sùunak-ech
t-u=láak’
ha’b=e’,)
SR:IRR REP=turn\ATP:SUBJ-B2SG PREP-A3=other year=TOP
túumben le=nah=o’
new(B3SG) DET=house=D2
‘(When you came here two years ago/when you return next (lit. the
other) year,) the house is/was/will be new’
The rationale behind this dichotomy in the grammatical treatment of propositions that concern the realization of events and propositions about states
deserves further attention.
The case against an anaphoric-tense analysis of the Yucatec AM markers rests on Occam’s Razor: the semantic contribution of a given marker in
a given utterance can be either relative tense or some modal or aspectual
meaning, but should not be assumed to be both except in the face of compelling evidence. Given this principle, the relative tense analysis of a marker
is defeated by demonstrating aspectual or modal meaning components.
Such meaning components are established on the basis of failure to entail
event realization (this applies to all modal AM markers, the imperfective,
progressive, and prospective aspect markers, and the remote and immediate
future markers) or incompatibility with event time adverbials (this applies
to the terminative aspect marker and the five AM markers expressing degrees of remoteness). The examples in (25)–(26) illustrate failure to entail
event realization for the progressive (25) and obligative (26) AM markers:
(25) Káa=h-tàal-ech
way h-ts’o’k
ka’=p’éel ha’b=e’,
CON=PRV-come-B2SG here PRV-end(B3SG) two=CL.IN year=D3
táan in=mèet-ik
le=nah=o’.
PROG A1SG=do:APP-INC(B3SG) DET=house=D2
Ba’x=e’ ma’ h-bèey=chah
what=TOP NEG PRV-thus=INCH.CMP(B3SG)=D4
Temporal anaphora in a tenseless language 29
in=ts’o’k-s-ik
tuméen h-k’oha’n-chah-en.
A1SG=end-CAUS-INC(B3SG) CAUSE PRV-sick-INCH.CMP-B1SG
‘When you came here two years ago, I was building the house. But, I
wasn’t able to finish it because I became ill.’
(26) Káa=h-tàal-ech
way h-ts’o’k
ka’=p’éel ha’b=e’,
CON=PRV-come-B2SG here PRV-end(B3SG) two=CL.IN year=D3
yan in=mèet-ik
ka’ch
le=nah=o’.
OBL A1SG=do:APP-INC(B3SG) formerly DET=house=D2
Ba’x=e’ ma’ h-bèey=chah=i’,
what=TOP NEG PRV-thus=INCH.CMP(B3SG)=D4
tuméen h-k’oha’n-chah-en.
CAUSE PRV-sick-INCH.CMP-B1SG
‘When you came here two years ago, I had to build the house. But, it
didn’t work out because I became ill.’
Incompatibility with event time adverbials has been exemplified for the
terminative AM marker in (23). The remote future marker has both diagnostic properties: it fails to entail event realization (27) and is incompatible
with event time adverbials (28).
(27) Bíin
in=mèet-∅
le=nah=o’,
ba’x=e’, ma’
REMF A1SG=do:APP-SUBJ(B3SG) DET=house=D2 what=TOP NEG
inw=ohel
wáah yan u=bèey-tal.
A1SG=knowledge(B3SG) ALT OBL A3=thus-INCH .INC
‘It will be a long time before I build the house, but I don’t know
whether it will be possible.’
(28) *Bíin in=mèet-∅
le=nah
REMF A1SG=do:APP-SUBJ(B3SG) DET=house
te=àanyo k-u=tàal=o’.
PREP=year IMPF-A3=come=D2
intended: ‘I will build the house next year.’
The AM markers of temporal distance deserve special attention in the context of the tenselessness analysis. As mentioned in section 2, these differ
from better studied “metrical tense” systems (cf., e.g., Comrie 1985: ch.4;
Dahl 1984, 1985: 120–128) in a number of respects. Most importantly, as
demonstrated here, they do not encode absolute tense. Neither is their use
30 Jürgen Bohnemeyer
obligatory for reference to an event at the specified distance from topic
time – they are used to emphasize the degree of remoteness much the same
way adverbials such as recently and a long time ago are in English. They
could be analyzed as optional anaphoric/relative metrical tenses. However,
as shown above, they are stative predicates that do not permit event time
specifications. Instead of specifying the distance between topic time and
coding time, as absolute metrical tenses would, or the distance between
topic time and some reference point, as relative metrical tenses would, they
specify the distance between topic time and the time of the event described
by the verbal core. They are thus not tenses in the sense of the definition
given in the beginning of this section: they do not constrain the topic time
of the utterance vis-à-vis coding time or some other reference point. Instead,
their semantics concerns the relation between topic time and event time,
much like that of the aspectual and modal AM markers.
Subordinate clauses and verbal cores can be divided into three classes: (a)
embedded verbal cores which occur as complements of matrix predicates
or adpositions – these are interpreted with respect to the topic time of the
matrix; (b) subordinate clauses which show the same aspect-mood marking
system as independent clauses – these are thus subject to the same argumentation advanced above for independent clauses; and (c) subordinate clauses
that show the reduced AM-marking systems briefly discussed in section 2.3.
As mentioned, these reduced systems are all governed by a realis-irrealis
mood contrast. In this case, the fact that the same irrealis form used for (absolute and relative) future time reference is also used for habitual and generic reference makes a tense analysis implausible from the start.
Yucatec does, of course, have means of explicitly constraining the topic
time of an utterance. Adverbials and subordinate clauses as illustrated in
most of the examples in this section will do the job just fine. This includes
two adverbs whose semantics is similar if not identical to that of present
and past tenses, respectively: the “topic time shifters” be’òora …=a’ ‘now’,
which restricts topic time to overlap with coding time, and ka’ch(il) ‘formerly’, which situates topic time in the past of utterance time. The latter is
illustrated in the second line of (26) above. These are not considered tenses
here because they are clearly not required by the morphosyntactic form of
the sentence – they are just adverbs that optionally occur in the same positions as a host of other adverbs. As a result, their pragmatics is also quite
different from that of tense marking in Indo-European languages. However,
it seems quite possible that optional tense markers in languages that have
them function similarly to the Yucatec topic time shifters. Be’òora …=a’
‘now’ is predominantly used to implicate that a state asserted to hold at
Temporal anaphora in a tenseless language 31
topic time did not hold at some relevant earlier time, and ka’ch(il) triggers
the inverse implicature that the state asserted to hold at topic time does not
hold any longer at coding time. For example, in (26), ka’ch(il) is used to
indicate that the state of “obligation” to build the house no longer holds at
utterance time. Of course, tense marking in English carries the same implicatures; the difference is that the Yucatec topic time shifters are used only
when this implicature is intended to be conveyed.
4. Temporal anaphora in Yucatec
Section 3 has summarized the evidence suggesting that Yucatec is a tenseless language. Tenses serve to constrain the topic time of an utterance to
the present, past, or future of coding time (in the case of absolute tense) or
some reference point (in the case of relative tense). It is safe to assume that
it is as important for Yucatec speakers as it is for English speakers to be able
to distinguish narrative accounts of past events from predictions of future
events or declarations of intentions about future events and, for example,
descriptions of habits and statements of general rules. The question in the
present section is not whether Yucatec speakers are able to infer that the
topic time of any given utterance lies in the present, past, or future of coding
time or some reference point; the question is how exactly they do this.
A standard claim made in discussions of tenselessness is that adverbials
can be used to compensate for the lack of tense markers. While certainly not
false, this is misleading to the extent that it suggests that adverbials are more
frequent or play a more important role in discourses of tenseless languages
than in those of tensed ones. At least as far as Yucatec is concerned, this is
not the case. The events narrated in Yucatec folk tales are not anymore anchored to a calendrical time scale than those of Hansel and Gretel (see below for an illustration), and if a Yucatec speaker wishes to convey that the
bus I have been waiting for has already left or that a house is on fire in another part of the village or that they are planning to get married, they are
perfectly able and in fact likely to do so without using any temporal adverbials.
The argument to be advanced in this section is that to determine the order
between the topic time of an utterance and its coding time, Yucatec speakers
rely on the same “mechanism” speakers of English and other better studied
languages rely on to determine topic times in context. It so happens that this
mechanism is not needed to determine the relation between topic times and
coding times in tensed languages such as English because these relations are
32 Jürgen Bohnemeyer
expressed by tense markers. However, the mechanism plays a key role in
determining temporal relations between clauses in connected speech in
tensed and tenseless languages alike. The mechanism in question is, of
course, that of temporal anaphora. Examples of temporal anaphora have
been presented throughout the previous sections. Two excerpts from larger
texts may help getting a flavor for the matter. Example (29) is a passage
from a demon story (discussed in detail in Bohnemeyer 2003).
(29) Le=òotsil máak=o’, káa=h-bin
te’l ich
DET=poor person=D2 CON=PRV-go(B3SG) there in
‘The poor man, he went (out) there to’
le=kòol=o’.
Ti’, bin, yàan
te=ka’nal=o’,
DET=clear\ATP=D2 there HS EXIST(B3SG) PREP:DET=high=D2
‘the milpa (swidden, lit. ‘clearance’). There he was, they say, up high
(i.e., in a tree),’
chéen káa=t-y=il-ah
only káa=PRV-A3=see-CMP (B3SG)
‘(and) he saw’
u=tíip’-il,
bin, le=ba’l=o’;
túun
tàal
A3=appear-INC HS DET=thing=D2 PROG:A3 come
‘the thing (i.e., the demon) appear, they say; it was coming’
The first clause describes an event of a man going to work in his milpa or
swidden. The next clause is stative and describes the man’s location up in a
tree. The topic time of this state description is inferred to follow that of the
first clause (a point to be commented on below). The tree is understood to
be either on the milpa or on the way there (we do not actually learn whether
the man ever reached his goal, and the first clause does not entail this, contrary to its gloss). The third clause, like the first, is an event description in
the perfective aspect, introduced by the connective káa briefly discussed in
section 2, which is characteristic of narrative discourse. It introduces the
man’s perception of the approaching demon. The topic time of this clause
is understood to be the same as that of the preceding state description: the
man is up in the tree and sees the demon from there. The fourth and final
clause features the progressive aspect marker. It refers to the demon’s approach. Its topic time is understood to be the same as that of the preceding
two clauses. The aspect markers stipulate that the perception event is included in this topic time, whereas the times of the man’s being up in the
tree and the demon’s approach contain the topic time.
Temporal anaphora in a tenseless language 33
In the second excerpt, the speaker talks about the hurricane Roxana
which hit his village in 1996 about a week before the recording.
(30)
Káa=h-k’uch-o’n
túun way te=kàah-il
CON=PRV-arrive-B1PL so.then here PREP:DET=live-REL
‘So then (when) we arrived here in the village’
x-Yaxley-il=e’,
k=il-ik=e’
tuláakal
F-Yaxley-REL=TOP IMPF:A1PL=see-INC(B3SG )=TOP all
‘of Yaxley, we saw (that) all’
máak=e’,
táan uy=a’l-ik-o’b=e’
person=TOP PROG A3=say-INC(B3SG)-3PL =TOP
‘people (i.e., everybody), they were saying’
hach ts’-uy-u’b-ik-o’b
ti’ ràadyo=e’
really TERM-A3=perceive-INC(B3SG)-3PL PREP radio=TOP
‘(that) they had really heard on the radio’
túun
tàal le=siklòon=o’
come DET=cyclone=D2
‘(that) the hurricane was coming’
PROG:A3
The first clause is marked for perfective aspect and introduces the event of
the speaker’s arrival in the village. As a “topic clause” (see section 2.3, 2.5),
it sets the topic time for the following main clause, which in fact is the
topic time for the entire passage. The main clause contains the perception
verb il ‘see’ marked for imperfective aspect. This combination is an idiom
frequently used in first-person narratives to describe the narrator’s realization of previously unknown facts. Semantically, this idiom is interpreted
perfectively. In (30), the realization is understood to take place at or after
the speaker’s arrival. The object of the speaker’s realization is described by
a sequence of three clauses. Semantically, each of these is coindexed with
the “theme” argument of the previous; but syntactically, they are not embedded as complements, but rather linked anaphorically in a kind of topic
chain. The first of these is marked for progressive aspect and talks about
what the villagers were saying at topic time, which is still the time of the
speaker’s arrival. Because of the progressive, topic time is understood to
fall into the time of the villagers saying this, so their talk is described as
having started before the speaker’s arrival. The final two clauses describe
what the villagers were saying: they had heard on the radio that the hurricane was indeed going to hit their area. The clause referring to the villagers
hearing the news on the radio carries the terminative AM marker, which
34 Jürgen Bohnemeyer
functions much like a (tenseless) perfect: topic time is presented as falling
into the post-state of the villagers hearing the news on the radio. In other
words, they are said to have heard it before the speaker’s arrival. The final
clause refers to the approach of the hurricane. It is marked for progressive
aspect, so the approach is presented as ongoing at topic time. It will be inferred that it was in fact already ongoing at the time this was announced on
the radio, but this is not strictly entailed in (30).
The similarity to the English examples discussed in the introduction is
intuitively obvious: some clauses introduce a new temporal perspective – in
the present framework, the topic time – perhaps advancing the previous one,
whereas others are interpreted with respect to this topic time. And the difference seems to depend, among other factors, on the aspectual properties
of the clauses.
The existence of temporal anaphora in a tenseless language such as Yucatec is of course not surprising (although Bohnemeyer 1998b for Yucatec
and Bittner (2008) for Kalaallisut are in fact the first descriptions of temporal anaphora in tenseless languages). Nevertheless, it is worth pointing out
that in Partee 1973 and in the DRT literature adopting the concept, temporal
anaphora is treated as part of the interpretation of tense markers in terms of
a variable they introduce whose value is determined in context. The evidence from Yucatec and Kalaallisut makes it clear that utterances are interpreted with respect to topic times whether or not they are marked for tense,
and that the determination of these topic times follows similar principles in
tensed and tenseless languages.
In order to sketch an informal account of the interaction between temporal anaphora and aspect-mood marking, I would like to introduce the notion
of the “Natural temporal reference point”:
(31) Natural temporal reference point (NTRP)
A time interval t is an NTRP in a given discourse iff t is identified in
that discourse as either (a) the coding time of some utterance or (b) a
calendrical time interval or (c) an event time (the “run time” of an
event described in the discourse).
This principle says that the times suitable as temporal reference points in
discourse are the times identified on some calendar or clock-time scale and
the times of events – including events described in the discourse and, in the
case of deictic reference, the event of the production and/or comprehension
of the utterance. What does not qualify a time interval as a suitable reference
point or NTRP is the fact that some event is in progress at this time or that
Temporal anaphora in a tenseless language 35
some state holds during it – for example a causal pre- or result state of some
event, a state characterizing the realization of an event in some possible
worlds, or a state characterizing the distance of the event from topic time. It
follows that only perfective clauses, but not non-perfective clauses, introduce NTRPs.19 Non-perfective clauses do not provide such reference points,
but on the contrary require them for their interpretation. This is not to say
that non-perfective clauses cannot be used to introduce topic time variables.
In fact, this happens whenever a non-perfective clause is used to provide
background information for a narrative sequence. An example is the stative
topic clause in the second line of (29). Stative clauses are strongly preferred
to be interpreted imperfectively. This means that the topic time introduced
by the stative clause in (29) is not the entire time the man spent up in the
tree, but rather some subinterval. If a perfective clause follows, that subinterval is understood to be a suitable time frame that contains the time of the
event described by the perfective clause. The stative clause introduces the
topic time variable for the perfective clause, but it is the inclusion of the
event time of the perfective clause in this topic time that determines its value
sufficiently to make it an NTRP. Suppose now instead of the perfective
clause describing the perception of the demon’s approach, the text would
continue with the final progressive clause of (29), which describes the demon’s approach as being in progress at topic time, as suggested in (32):
(32) Ti’, bin, yàan
te=ka’nal=o’,
there HS EXIST(B3SG) PREP:DET=high=D2
‘There he was, they say, up high (i.e., in a tree),’
le=ba’l=o’;
túun
tàal
DET=thing=D2 PROG:A3 come
‘the thing (i.e., the demon), it was coming’
The discourse in (32) might serve as background for a third clause using
the perfective to place some event into the topic time of which we so far
know that it falls into both the time of the man’s being in the tree and the
time of the demon’s approach being in progress. But as a self-contained
episode description, (32) not only makes a poor narrative (“nothing ever
19
In non-narrative discourses, clauses formed with the imperfective AM marker
and irrealis topic clauses may be used to introduce new reference points. Examples of the latter option are (10) and (19) above. It seems that in these cases,
event realization and thus the introduction of NTRPs is treated inside the scope
of modal or habitual/generic operators.
36 Jürgen Bohnemeyer
happens!”); it is also difficult to interpret. In (32), the imperfectively interpreted stative clause is combined with a progressive clause, which is semantically likewise imperfective. The progressive clause requires selection
of a suitable topic time interval that falls into the time of the demon’s approach. Even assuming the two topic times are identical, neither clause
provides a suitable reference point for its resolution. Are the two states of
affairs coextensive? Was the run time of one included in that of the other?
Or did the two overlap partially? Principle (33) formulates the role of
NTRPs in the selection of topic times:
(33) Preferred topic time selection
The topic times selected in a given discourse context are preferred to
be identical to or include NTRPs identified in the same discourse
context.
The possibility of inclusion of the NTRP in the topic time is mentioned in
(33) because the topic times of perfective clauses are assumed in the present
framework to contain the run times of the events described by the clauses
(see Figure 1 in section 2.3). Principle (33) can be understood as a constraint
on coherent discourses. Speakers craft their discourses so as to satisfy this
constraint. The means at their disposal to manipulate topic time selection
vary somewhat from language to language. In Yucatec, these are mainly
viewpoint aspect and modal operators and adverbials; in English and other
Indo-European languages, they also include tenses and temporal connective
constructions.
Principles (31) and (33) together account for two key differences in the
discourse behavior of perfective and non-perfective clauses. Non-perfective
clauses trigger what may be called binding implicatures to the effect that
their topic times are identical to some salient NTRP accessible in context.
This can be coding time, giving rise to deictic interpretations, the time
specified by some calendrical adverbial, or the event time of a perfective
clause in surrounding discourse. Perfective clauses trigger no such implicatures because their topic times already include NTRPs – the run times of
the events they describe. Perfective clauses can be used to introduce reference points “binding” the topic times of non-perfective clauses, and in the
context of other perfective clauses, they may trigger the well-known referential shift interpretations. Consider (34):
Temporal anaphora in a tenseless language 37
(34) Káa=h-tàal-ech
way h-ts’o’k
ka’-p’éel ha’b=e’,…
CON=PRV-come-B2SG here PRV-end(B3SG) two-CL.IN year=TOP
‘(When) you came here two years ago,…’
a. …káa=t-in=mèet-ah
le=nah=o’
CON=PRV-A1SG =do:APP-CMP(B3SG) DET=house=D2
‘…I built the house’
b. …táan in=mèet-ik
le=nah=o’
PROG A1SG=do:APP-INC(B3SG) DET=house=D2
‘…I was building the house’
c. …ts’o’k in=mèet-ik
le=nah=o’
TERM
A1SG=do:APP-INC(B3SG) DET=house=D2
‘…I had built the house’
d. …mukah in=mèet
le=nah=o’
PROSP
A1SG=do:APP(SUBJ)(B3SG ) DET=house=D2
‘…I was going to build the house’
The clauses in (34a–d) follow the same perfective topic clause in the first
line of (34). The perfective clause in (34a) is interpreted with respect to a
new, “shifted” topic time not identical to the event time of the topic clause.
Pragmatic inferences to be discussed below will locate the former just after
the latter, giving rise to the interpretation that the speaker started building
the house upon the addressee’s arrival. In contrast, the continuations using
progressive, terminative, and prospective aspect markers in (34 b–d) are
understood with respect to the event time of the topic clause as their topic
time, giving rise to the interpretations that the construction was in progress
(b), completed (c), or being planned (d) at the time of the addressee’s arrival.
Binding implicatures are stereotype implicatures of the kind discussed
by Atlas & Levinson 1981, generated by Grice’s second Quantity maxim
(“Do not make your contribution more informative than is required”).20 Ref20
If perfective clauses triggered binding implicatures, too, this would make narrative progression by referential shift impossible. Narratives would then require
every perfective clause to be equipped with some device – an adverbial or some
other expression – that explicitly signals “updating” of the topic time variable.
As a matter of fact, this is precisely the function of the connective káa in the examples above, as discussed in section 2.5. The reason the perfective AM marker
of Yucatec is accompanied by this connective in narrative discourse is that it is
semantically vague between proper perfective (event time included in topic time)
and perfect-like result-state interpretations (topic time included in the time of
the result state).
38 Jürgen Bohnemeyer
erential shift, on the other hand, is the product of a combination of the failure to trigger binding implicatures and another stereotype implicature, this
time to iconicity, i.e., to the effect that the order of clauses iconically reflects
the order of events. With non-perfective clauses, this iconicity implicature
is overridden by the binding implicature, since the latter is more specific –
it yields “binding” of the topic time by a specific contextually accessible
NTRP, whereas the iconicity implicature is satisfied by any time interval
following the topic time of the preceding clause.
As generalized conversational implicatures, the binding and iconicity
implicatures are defeasible default interpretations (Levinson 2000) triggered
by the use of viewpoint-aspectual and modal operators in suitable contexts.
The stative clause in (29) in fact illustrates blocking of the binding implicature: its topic time is interpreted to be shifted vis-à-vis that of the initial
perfective clause because encyclopedic knowledge suggests that the man
cannot have been sitting in a tree at the time he left for his milpa. A similar
effect is illustrated in (35): the binding implicature is blocked because a
balloon cannot be continued to be inflated once it has burst.
(35) Táan u=p’uru’s-t-ik=e’,
káa=h-xíik-ih
PROG A3=inflate-APP-INC (B3SG)=D3 CON=PRV-burst-CMP (B3SG)
‘She was inflating (the balloon), (when) it burst’
Defeasibility of the iconicity implicature can be illustrated by (12)–(13) in
section 2.5, repeated here for convenience:
(36) Pedro=e’ káa=t-u=ts’íib-t-ah
hun-p’éel
Pedro=TOP CON=PRV-A3=write-APP-CMP (B3SG) one-CL.IN
kàarta=e’, káa=t-u=ts’u’uts’-ah
hun-p’éel chamal
letter=TOP CON=PRV-A3=suck-CMP (B3SG) one-CL.IN cigarette
‘Pedro, (when) he wrote a letter, he smoked a cigarette’
(preferred interpretation sequential)
(37) Káa=t-u=ts’íib-t-ah
hun-p’éel kàarta
CON=PRV-A3=write-APP-CMP (B3SG) one-CL.IN letter
Pedro=e’, Juán=e’, káa=t-u=ts’u’uts’-ah
Pedro=TOP Juán=TOP CON=PRV-A3 suck-CMP (B3SG)
hun-p’éel chamal
one-CL.IN cigarette
‘(When) Pedro wrote a letter, Juán smoked a cigarette’
(preferred interpretation for most consultants simultaneous)
Temporal anaphora in a tenseless language 39
The sequence of two perfective clauses in (36) triggers the familiar shift
interpretation, while the one in (37) fails to do so because the two actions
described have different agents. The discourse in (37) is certainly compatible with a sequential interpretation; it’s just that out of context, most speakers consider the interpretation according to which the two actions occurred
at the same time more salient. Bittner (2008), in her analysis of temporal
anaphora in Kalaallisut within her “online update” framework, claims that
temporal anaphora is in fact monotonic in “aspectually fully explicit” languages such as Kalaallisut. The examples presented above suggest that this
analysis does not apply to Yucatec, even though Yucatec may well be considered “aspectually fully explicit”, at least as far as the grammar of event
descriptions, as opposed to state descriptions, is concerned.
The Gricean analysis of the inferences involved in aspect-based temporal
anaphora resolution has the advantage over the standard DRT account that
it offers an explanation for why these inferences and no others are triggered
by aspectual operators and that it predicts the conditions under which these
inferences are cancelled or blocked. The standard DRT approach accommodates the defeasibility of aspect-based temporal anaphora resolution as
vagueness, based on the “event structure” model of Kamp 1979: the temporal reference of utterances is interpreted with respect to “instants”, which in
turn are constituted by sets of pairwise overlapping events. Thus, for example, while (34a) is said to introduce a new reference point at an instant that
follows the reference point of the first clause, whereas (34b) is interpreted
with respect to the reference point of the first clause, the two events may in
both cases either overlap or follow one another. What this approach fails to
explain is the default character of the inferences: sequential ordering in
(34a) and simultaneity in (34b) will be inferred unless these interpretations
are blocked or cancelled in context, as in (37) and (35). This default character follows from aspect-based temporal anaphora being rooted in generalized conversational implicatures, i.e., utterance-type meanings.
The account of temporal anaphora sketched above straightforwardly
generalizes to the deictic uses of the aspect-mood markers. The binding
implicatures triggered by non-perfective clauses are satisfied by calendrical
adverbials or salient reference points available in context; where these are
absent coding time takes over as the NTRP in accordance with (31) and
(33). This gives rise to the deictic interpretations discussed in section 3:
topic time is (or includes) coding time and is itself included in the run time
of the event under description (with the progressive AM marker), a result
state (with the terminative AM marker), a pre-state (with the prospective
AM marker), some state that characterizes the realization of the described
40 Jürgen Bohnemeyer
event in possible worlds (with the modal AM markers), or some state that
characterizes the distance of the described event from topic time (with the
“metrical” AM markers). Smith, Perkins, & Fernald 2007 note the same
affinity of non-perfective clauses for deictic interpretations in Navajo, but
propose a special “Deictic Principle” to account for it. In the present treatment, this affinity follows from the fact that coding time is a natural temporal reference point in combination with the semantics of non-perfective aspect-mood markers and the generalized conversational implicatures that
govern temporal anaphora.
Clauses formed with the perfective AM marker are likewise used with
topic time set to coding time, under result-state interpretations. Under perfective interpretations, they are excluded from deictic uses except for the
marginal “blow-by-blow” online narrative context. This follows from the
fact that the topic times of semantically perfective clauses include the run
times of the events described by these clauses; these cannot be included in
coding time except in the “blow-by-blow” scenario where coding time and
topic time are continuously updated. In other words, all of deictic future
and past time reference in Yucatec in fact involves present topic times
(similarly Bittner 2005 for Kalaallisut).
5. Concluding remarks
Why do some languages have tense marking whereas others lack it? In my
opinion, for essentially the same reason some languages (including English,
Ewe, German, and Yucatec) mark their noun phrases for definiteness while
the speakers of others (Estonian, Latin, Mandarin, Russian, Korean, and
many more) seem to get by just fine without this device. Similarly, in some
languages, noun phrases are marked for noun class or gender or trigger
noun class or gender agreement (e.g., German, Kinyarwanda, Latin, and
Russian) whereas gender and noun class play no role in the functional category system of other languages (e.g., in English, Estonian, Ewe, and Yucatec). In many languages, event descriptions are obligatorily marked for
their viewpoint aspect (e.g., English, Ewe, Russian, Yucatec); but in some,
they are not (e.g., Estonian and German). In languages such as Quechua and
Turkish, clauses are obligatorily marked for the source of information the
speaker purports to rely on, whereas in all other languages mentioned above,
this is merely optionally indicated by lexical means. The reason for this kind
of crosslinguistic variation in the functional category system seems to be
that the expression of functional categories such as tense, viewpoint aspect,
Temporal anaphora in a tenseless language 41
definiteness, gender, noun class, and evidentiality is not necessary for conveying the intended communicative content of linguistic utterances. The
relevant conceptual distinctions are made whether or not they are expressed
linguistically and speakers can rely on pragmatic means to communicate
them where needed. Where functional categories are expressed, they serve
to disambiguate and to facilitate reference resolution. One can thus surmise
that there is a certain division of labor between pragmatics and the functional category system and a tradeoff between expressed and unexpressed
categories. The expression of a rich system of aspectual and modal distinctions and simultaneous absence of tense marking in Yucatec exemplifies
this tradeoff.
Acknowledgements
I am indebted to my Maya teachers and consultants, above all to Sebastián
Baas May, Vicente Ek Catzin, Fulgencio Ek Ek, Amalia Ek Falcón, Máximo
Ek Pat, Ernesto May Balam, Ramón May Cupul, Antonio May Ek, Saturnino
May Ek, Norma May Pool, and Justina Pat May.
References
Ayres, Glenn and Barbara Pfeiler
1997
Los Verbos Mayas: La Conjugación en el Maya Yucateco Moderno
[Mayan Verbs: Conjugation in Modern Yucatec Maya]. Mérida, Yucatan: Ediciones de la Universidad Autónoma de Yucatán.
Abusch, Dorit
1985
On verbs and times. Doctoral dissertation, University of Massachusetts, Amherst.
Bertinetto, Pier Marco and Mario Squartini
1995
An attempt at defining the class of ‘gradual completion verbs’. In
Temporal reference. Vol. 1: Semantic and syntactic perspectives, P.
M. Bertinetto, V. Bianchi, J. Higginbotham, and M. Squartini (eds.),
11–26. Torino: Rosenberg & Sellier.
Bittner, Maria
2005
Future discourse in a tenseless language. Journal of Semantics 22:
339–387.
2007
???
42 Jürgen Bohnemeyer
2008
Aspectual universals of temporal anaphora. In Theoretical and Crosslinguistic Approaches to the Semantics of Aspect, S. Rothstein (ed.).
Amsterdam: John Benjamins.
Blair, Robert W. and Refugio Vermont Salas
1965–67 Spoken (Yucatec) Maya. Chicago: University of Chicago, Department
of Anthropology.
Bohnemeyer, Jürgen
1998a
Temporal reference from a Radical Pragmatics perspective: Why
Yukatek does not need to express ‘after’ and ‘before’. Cognitive Linguistics 9 (3): 239–282.
1998b
Time relations in discourse: Evidence from Yukatek Maya. Doctoral
dissertation, Tilburg University.
1998c
Die Stellung sententialer Topics im Yukatekischen (The position of
sentential topics in Yucatec). In Deskriptive Grammatik und allgemeiner Sprachvergleich [Descriptive grammar and general language
comparison], D. Zaefferer (ed.), 55–85. Tübingen: Niemeyer.
2000a
Event order in language and cognition. In Linguistics in the Netherlands 17, H. de Hoop and T. van der Wouden (eds.), 1–16. Amsterdam:
John Benjamins.
2000b
Where do pragmatic meanings come from? The source of temporal
inferences in discourse coherence. In Samenhang in diversiteit. Opstellen voor Leo Noordman [Unity in diversity. Papers presented to
Leo Noordman], W. Spooren, T. Sanders, and C. van Wijk (eds.),
137–153. Tilburg: Tilburg University.
2001
Argument and Event Structure in Yukatek Verb Classes. In The Proceedings of SULA: The Semantics of Under-Represented Languages
in the Americas, J.-Y. Kim and A. Weerle (eds.), 8 –19. Amherst, MA:
GLSA (University of Massachusetts Occasional Papers in Linguistics
25).
2002
The grammar of time reference in Yukatek Maya. Munich: Lincom
Europa.
2003
Invisible time lines in the fabric of events: Temporal coherence in
Yucatec narratives. Journal of Linguistic Anthropology 13 (2): 139–
162.
2004
Split intransitivity, linking, and lexical representation: the case of
Yukatek Maya. Linguistics 42 (1): 67–107.
2007
Morpholexical Transparency and the argument structure of verbs of
cutting and breaking. Cognitive Linguistics 18 (2): 153–177.
in press Linking without grammatical relations in Yucatec: Alignment, extraction, and control. In Issues in functional-typological linguistics and
language theory, Y. Nishina, Y. M. Shin, S. Skopeteas, E. Verhoeven,
and J. Helmbrecht (eds.). Berlin /New York: Mouton de Gruyter.
Temporal anaphora in a tenseless language 43
Bohnemeyer, Jürgen and Penelope Brown
2007
Standing divided: Dispositionals and locative predications in two
Mayan languages. Linguistics 45 (5–6): 1105–1151.
Bohnemeyer, Jürgen and Mary D. Swift
2004
Event realization and default aspect. Linguistics and Philosophy 27
(3): 263–296.
2006
Force dynamics and the progressive. Annual Meeting of the Linguistic Society of America; Albuquerque, NM.
Bricker, Victoria R.
1979
Wh-questions, relativization, and clefting in Yucatec Maya. In Papers
in Mayan linguistics 3, L. Martin (ed.), 107–136. Columbia, MO:
Lucas Brothers.
Comrie, Bernard
1976
Aspect. Cambridge: Cambridge University Press.
1985
Tense. Cambridge: Cambridge University Press.
Dahl, Östen
1984
Temporal distance: Remoteness distinctions in tense-aspect systems.
In Explanations for language universals, B. Butterworth, B. Comrie,
and Ö. Dahl (eds.), 105–122. Berlin /New York: Mouton de Gruyter.
1985
Tense and aspect systems. Oxford: Blackwell.
Dowty, David R.
1979
Word meaning and Montague Grammar. Dordrecht: Reidel.
Ethnologue
2005
Freed, A.
1979
The semantics of English aspectual complementation. Dordrecht:
Reidel.
Gutiérrez Bravo, Rodrigo
2006
Foco de agente y cláusulas relativas en maya yucateco [Agent focus
and relative clauses in Yucatec Maya]. Manuscript, CIESAS-DF.
Hay, Jennifer, Christopher Kennedy, and Beth Levin
1999
Scalar structure underlies telicity in ‘degree achievements’. In SALT
IX, T. Mathews & D. Strolovitch (eds.), 127–144. Ithaca: CLC Publications.
Hinrichs, Erhard
1981
Temporale Anaphora im Englischen. M.A. dissertation, University
of Tübingen.
1986
Temporal anaphora in discourses of English. Linguistics and Philosophy 9: 63–82.
Horn, Laurence R.
1972
On the semantic properties of logical operators in English. Mimeo.
Bloomington, IN: Indiana University Linguistics Club.
44 Jürgen Bohnemeyer
Jackendoff, Ray S.
1983
Semantics and cognition. Cambridge, MA: MIT Press.
Kamp, Hans
1979
Events, instants, and temporal reference. In Semantics from different
points of view, R. Bäuerle, U. Egli, and A. von Stechow (eds.), 376–
418. Berlin: Springer.
Kamp, Hans and Uwe Reyle
1993
From discourse to logic: Introduction to modeltheoretic semantics of
natural language, formal logic and Discourse Representation Theory.
Dordrecht: Kluwer.
Kamp, Hans and Christian Rohrer
1983
Tense in Texts. In Meaning, use, and interpretation of language, R.
Bäuerle, C. Schwartze, and A. von Stechow (eds.), 250–269. Berlin /
New York: Mouton de Gruyter.
Klein, Wolfgang
1994
Time in language. London: Routledge.
2006
Kratzer, Angelika
1977
What ‘must’ and ‘can’ must and can mean. Linguistics and Philosophy 1: 337–355.
1981
1991
Krifka, Manfred
1992
Thematic relations as links between nominal reference and temporal
constitution. In Lexical matters, I. A. Sag and A. Szabolcsi (eds.), 29–
54. Menlo Park, CA: CSLI /SRI International.
Landman, Fred
1992
The progressive. Natural Language Semantics 1: 1–32.
Levin, Beth
1993
English verb classes and alternations. Chicago, IL: University of Chicago Press.
Levin, Beth and Malka Rappaport Hovav
1995
Unaccusativity. Cambridge, MA: MIT Press.
Levinson, Stephen C.
2000
Presumptive meanings. Cambridge, MA: MIT Press.
Li, Charles N. and Sandra A. Thompson
1981
Mandarin Chinese: A functional reference grammar. Berkeley, CA:
University of California Press.
Lucy, John A.
1994
The role of semantic value in lexical comparison. Linguistics 32:
623–656.
Temporal anaphora in a tenseless language 45
Partee, Barbara
1973
Some structural analogies between tenses and pronouns in English.
The Journal of Philosophy 70: 601–609.
1984
Nominal and temporal anaphora. Linguistics and Philosophy 7: 243–
86.
Portner, Paul
1998
The progressive in modal semantics. Language 74: 760 –787.
Reichenbach, Hans
1947
Elements of Symbolic Logic. New York, NY: Free Press.
Shaer, Benjamin
2003
Toward the tenseless analysis of a tenseless language. J. Anderssen,
P. Menéndez Benito, and A. Werle (eds.), 139–56. Proceedings of
SULA 2. Amherst, MA: GLSA.
Smith, Carlota S.
1978
Jespersen’s ‘move and change’ class and causative verbs in English.
In Linguistic and Literary Studies in Honor of Archibald A. Hill, Vol.
2: Descriptive Linguistics, M. A. Jazayery, E. C. Palome, and W.
Winter (eds.), 101–109. The Hague: Mouton.
Smith, Carlota S., Ellavina T. Perkins, and Theodore B. Fernald
2007
Time in Navajo: Direct and indirect interpretation. International
Journal of American Linguistics 73 (1): 40–71.
Tonhauser, Judith
2003
F-constructions in Yucatec Maya. In Proceedings of SULA 2, J. Anderssen, P. Menéndez Benito, and A. Werle (eds.), 203–223. Amherst,
MA: GLSA.
Van Valin, Robert D. Jr. and Randy J. LaPolla
1997
Syntax. Cambridge: Cambridge University Press.
Vapnarsky, Valentina
1995
La interpretación de las señas divinas (chiikul) entre los mayas yucatecos de Quintana Roo. In Antropología simbólica, Marie-Odile Marion (ed.). México D.F.: ENAH.
1999
Expressions et conceptions de la temporalité chez les mayas yucateques (Mexique) (Expressions and concepts of temporality among the
Yucatec Mayans [of Mexico]). Ph.D. dissertation, Department of Ethnology and Comparative Sociology, Université de Paris X Nanterre.